Healthcare

PocketDoc

A phone-to-laptop relay bringing offline medical triage - Whisper, RAG and an LLM, no cloud

transcriptionragvision

A BeagleY-AI board is too underpowered to run QVAC's native ML binaries directly. PocketDoc turns that limitation into the architecture: the board becomes a lightweight web gateway a phone connects to over WiFi, while a laptop on the same network does all the actual AI work - MedPsy-4B, Whisper STT, and a RAG pipeline - through the QVAC SDK.

Phone, board, laptop

A phone's browser talks to the board over WebSocket/HTTP; the board relays that text, voice recording, or camera photo to the laptop over the LAN; the laptop runs the actual inference and streams a response back the same way. Camera photos of symptoms are read by MedPsy-4B's multimodal path, voice queries go through Whisper Tiny first, and every query is grounded by RAG over a local medical knowledge base.

architecture (README.md)

📱 Phone/Browser  <--WiFi (WebSocket/HTTP)-->  🔧 BeagleY-AI (Board Server, :3000)
                                                          │ HTTP / SSE (LAN)
                                                          ▼
                                              💻 Laptop (AI Provider, :4000)
                                              QVAC SDK: MedPsy-4B LLM ·
                                              Whisper Tiny STT · Gemma
                                              Embeddings · RAG Pipeline

Demo video

Watch the PocketDoc demo →

Why a $50 board can run a medical triage assistant

The BeagleY-AI board cannot run QVAC natively, so it acts as a thin web gateway instead
All AI inference - LLM, Whisper STT, RAG - runs on a laptop reached over the local WiFi network
Text, voice, and camera queries all resolve through the same on-device MedPsy-4B and RAG pipeline

Ship your own local AI

One install in JavaScript or Python. 10+ AI tasks. Private, offline and free - on the hardware you already own.

Read the docs